An online monitoring system for temperature of whole cold chain based on internet of things

Abstract

The application provides a full cold chain temperature online monitoring system based on Internet of Things, comprising: a multi-source sensing layer arranged at a cold chain storage node, a cold chain transportation carrier and a terminal distribution node; the multi-source sensing layer comprises a multi-point temperature sensor array, a position sensor and an environment sensor; the multi-point temperature sensor array is arranged in a distributed manner at different depths and positions of a space to form a three-dimensional monitoring network of a space temperature field; the multi-source sensing layer sends collected temperature data to an edge computing gateway through a LoRa wireless communication module and sends the temperature data to a mobile terminal through a BLE communication module; the application can cover the whole cold chain, realize high-reliable temperature data collection and transmission, has intelligent early warning capability and data traceability.

Description

Technical Field

[0001] This invention relates to the field of cold chain logistics monitoring technology, specifically to an online temperature monitoring system for the entire cold chain based on the Internet of Things. Background Technology

[0002] Cold chain logistics is a crucial link in ensuring the quality and safety of temperature-sensitive goods such as food, pharmaceuticals, and biological products. The core requirements of the cold chain lie in the controllability and traceability of temperature throughout the entire process; any temperature deviation at any stage can lead to a decline in product quality, economic losses, or even safety incidents. With the continued expansion of the global cold chain logistics market and increasingly stringent food safety and pharmaceutical regulations, the industry's demand for end-to-end temperature monitoring of the cold chain is increasing dramatically.

[0003] Currently, the existing technology has the following main problems: First, the monitoring process is fragmented and lacks full-chain coverage. Existing systems are mostly designed for a specific link in the cold chain (such as warehousing or transportation), resulting in serious data silos and making it difficult to achieve real-time monitoring of the entire process from warehousing and transportation to last-mile delivery.

[0004] Secondly, temperature monitoring has blind spots. Traditional solutions rely on single temperature sensors at fixed locations, which cannot comprehensively reflect the temperature field distribution inside refrigerated trucks or cold storage facilities, resulting in the inability to detect localized temperature anomalies in a timely manner. Although some systems employ multi-point temperature acquisition, they lack efficient network management and redundancy verification mechanisms for the sensor network, leading to insufficient data reliability.

[0005] Third, data transmission and processing capabilities are insufficient. During cold chain transportation, vehicles are often in motion, and traditional WiFi networks suffer severe signal attenuation inside metal compartments. Bluetooth network coverage is also limited, resulting in high packet loss rates and unreliable transmission. Simultaneously, large amounts of raw data are directly uploaded to the cloud, leading to high network bandwidth pressure and high system response latency.

[0006] Fourth, there is a lack of intelligent early warning and traceability capabilities. Most systems only provide simple threshold alarm functions, lacking intelligent prediction of temperature change trends and dynamic adaptation to environmental changes, thus failing to achieve the transformation from "post-event traceability" to "pre-event early warning." Temperature alarm thresholds are usually fixed values, unable to be dynamically adjusted according to drastic changes in the external environment, leading to delayed warnings or frequent false alarms. Furthermore, temperature data lacks anti-tampering mechanisms during multi-party data transfer, making data reliability difficult to guarantee. Summary of the Invention

[0007] In view of this, the purpose of this invention is to provide an Internet of Things-based online temperature monitoring system for the entire cold chain, which can achieve high-reliability temperature data acquisition and transmission, intelligent early warning capabilities, and data traceability.

[0008] To achieve the above objectives, the present invention provides the following technical solution: This invention provides an Internet of Things-based online temperature monitoring system for the entire cold chain, which includes a multi-source sensing layer, an edge computing gateway, a cloud platform layer, and a user application layer.

[0009] The multi-source sensing layer is deployed at cold chain storage nodes, cold chain transportation carriers, and terminal delivery nodes, including multi-point temperature sensor arrays, position sensors, and environmental sensors.

[0010] The multi-point temperature sensor array is deployed in a distributed manner at different depths and orientations within the space, forming a three-dimensional monitoring network for the spatial temperature field. Specifically, the multi-point temperature sensor array includes at least three temperature sensor nodes, deployed at different depths and orientations within the space, for example, in a refrigerated truck compartment, deployed in three areas (front, middle, and rear) and at two heights (top and bottom). The temperature sensor nodes use high-precision digital temperature sensors with a measurement accuracy of no less than ±0.5℃, a resolution of no less than 0.1℃, and a sampling frequency configurable from 1Hz to 1 time / minute to meet the monitoring needs of different cold chain scenarios.

[0011] The environmental sensors include at least one of a humidity sensor, a door magnetic sensor, and a vibration sensor. The humidity sensor monitors humidity data in the cold chain environment, the door magnetic sensor monitors the opening and closing status and duration of doors in the space, and the vibration sensor monitors impact vibration events exceeding a preset threshold during transportation. These environmental sensors supplement key environmental information beyond temperature, providing multi-dimensional data support for comprehensively judging the cold chain status.

[0012] The multi-source sensing layer transmits the collected temperature data to the edge computing gateway via a LoRa wireless communication module and to the mobile terminal via a BLE communication module. The LoRa wireless communication module employs spread spectrum modulation technology to establish a wireless connection with the edge computing gateway in enclosed metal compartments or cold storage environments, achieving a data packet loss rate of less than 0.3%, effectively overcoming the severe signal attenuation problem of traditional wireless technologies in metal environments. The BLE communication module is used for short-range data synchronization with the handheld mobile terminal during inspections or loading / unloading, enabling offline data retransmission and on-site parameter configuration, ensuring data integrity in scenarios without public network coverage.

[0013] Edge computing gateways are deployed on cold chain transport vehicles or cold chain storage nodes to receive temperature data sent by multi-source sensing layers. An edge computing gateway includes a data preprocessing module, an anomaly detection module, a local decision-making module, a multi-protocol compatibility module, and a sensor data verification module.

[0014] The data preprocessing module performs noise reduction, outlier removal, and data compression on the received raw temperature data. For data compression, the preprocessing module uses a rotating door compression algorithm on the raw sensor data. By setting a compression deviation threshold, the continuously collected time-series data points are approximated as a sequence of line segments, and only the endpoint data of the line segments are uploaded, achieving a compression rate of at least 80% for the raw data. This significantly reduces network bandwidth load and cloud data storage pressure.

[0015] After receiving multiple temperature data points from a multi-point temperature sensor array, the sensor data verification module performs the following verification and fusion steps: First, it sorts the temperature values ​​collected from each sensor at the same time in ascending order and calculates the mean and standard deviation of the dataset. Second, based on the preset confidence probability and the current number of valid sensors, it obtains the critical value of the Grubbs criterion by looking up a table. Third, it calculates the Grubbs statistics for the maximum and minimum values ​​in the sorted data, compares the statistics with the critical value, and if the values ​​exceed the critical value, the corresponding data is identified as an outlier and removed. Fourth, it repeats the above steps until no data is identified as an outlier. Fifth, it calculates the arithmetic mean of the retained valid temperature data as the fused temperature value for that moment. This verification and fusion process effectively removes outlier data caused by individual sensor malfunctions or local interference, improving the overall reliability of the temperature data.

[0016] The anomaly detection module has a built-in adaptive threshold algorithm. This algorithm presets a basic temperature safety range based on the type of goods being transported and the current cold chain link, and acquires external temperature data collected by environmental sensors in real time to calculate the external temperature change rate. When the change rate exceeds a preset threshold, it automatically narrows the upper and lower limits of the safety range proportionally to dynamically adjust the temperature alarm threshold.

[0017] Specifically, the adaptive threshold algorithm performs the following process when dynamically adjusting the temperature alarm threshold: The anomaly detection module determines whether the current stage is warehousing, transportation, or distribution through the position sensor and the preset cold chain link logic; it reads the standard temperature safety lower and upper limits of the transported goods from the goods information synchronized from the cloud platform layer; it receives the external ambient temperature collected by the environmental sensor in real time and calculates the slope of the change of the current external temperature within the sliding time window; when the absolute value of the change slope exceeds the preset slope threshold, the upper and lower limits of the safety range are narrowed according to the magnitude of the slope by a proportional coefficient. The proportional coefficient is determined by a pre-calibrated goods temperature rise sensitivity curve, so that the alarm threshold is narrowed in advance when the external heat load increases sharply, thus achieving early warning.

[0018] When the local decision-making module detects an abnormal temperature, it immediately triggers a local alarm and automatically adjusts the cooling equipment, keeping the response delay to the millisecond level, ensuring that rapid local decision-making can still be achieved in the event of communication interruption or cloud response delay.

[0019] Furthermore, the edge computing gateway incorporates a lightweight AI model. This model takes preprocessed temperature sequences and chiller unit operating parameters as input, employs temporal convolutional networks or long short-term memory networks to extract temporal features of temperature changes, and combines this with historical fault label data to train anomaly detection and fault classifiers. By analyzing temperature change trends and historical sensor data, it identifies abnormal temperature patterns, enabling the prediction of potential chiller unit faults with an advance warning time of at least 48 hours. By deploying AI inference at the edge, the system can perform fault prediction without relying on the cloud, significantly reducing response latency while ensuring continuous monitoring capabilities even in offline environments.

[0020] The cloud platform layer interacts with the edge computing gateway via a 4G / 5G communication module, receiving temperature data that has been preprocessed at the edge. The cloud platform layer includes a time-series database engine, an intelligent analysis module, a digital twin module, a blockchain evidence storage module, and a collaborative scheduling module.

[0021] The intelligent analysis module includes a temperature trend prediction unit and a multi-level early warning unit.

[0022] The temperature trend prediction unit uses historical temperature sequences from the time-series database engine and external meteorological forecast data to predict temperature change trends over a preset time period using either an ARIMA or Prophet model. During the prediction process, the unit automatically performs model parameter order determination and seasonal decomposition, fully considering the autocorrelation, periodicity, and influence of external meteorological factors in the temperature sequence, thereby achieving accurate temperature change predictions.

[0023] The multi-level early warning unit adopts the following multi-level early warning strategy: Level 1, when the real-time collected temperature data exceeds the preset basic temperature safety range, a Level 1 alarm is triggered, indicating that the current temperature has deviated from the normal range and needs immediate attention; Level 2, when the rate of change of temperature data per unit time exceeds the dynamic change rate threshold set according to the thermal sensitivity characteristics of the goods, a Level 2 early warning is triggered, indicating that the temperature is rapidly deteriorating; Level 3, when the temperature trend prediction unit predicts that the temperature will exceed the safety range in the future, a Level 3 early warning is triggered, realizing the transformation from post-event response to pre-event prevention.

[0024] The digital twin module constructs a 3D digital model of the cold chain transport vehicle or storage space in the cloud, mapping the temperature field distribution collected and fused by a multi-point temperature sensor array in real time, and displaying the temperature distribution status in the 3D digital model in the form of a heat map. When the edge computing gateway or intelligent analysis module detects a local temperature anomaly, the digital twin module visually highlights the abnormal area in the 3D model and generates handling suggestions based on a preset expert rule base, including but not limited to adjusting the refrigeration setpoint, checking the integrity of door seals, and guiding the vehicle to the nearest dock for inspection, to help managers quickly locate the problem and take effective measures.

[0025] The blockchain-based evidence storage module performs hash calculations on key temperature data throughout the cold chain process and stores it on the blockchain. This key temperature data includes temperature records at each stage, records of abnormal events, and handover confirmation information. The module calculates a hash value for each piece of key data, packages the hash value into a blockchain transaction, and returns an evidence storage identifier. A QR code is generated based on this identifier, allowing consumers and regulatory agencies to scan the code to verify the tamper-proof cold chain data, thereby verifying the temperature compliance of goods throughout the cold chain process and enhancing data credibility and regulatory transparency.

[0026] The collaborative scheduling module is used to handle situations where abnormal temperatures cannot be restored through local adjustments. When a persistent temperature anomaly is detected in a cold chain transport vehicle and the local decision-making module is unable to restore the temperature to a safe range by adjusting the refrigeration equipment, the collaborative scheduling module obtains the real-time location of the abnormal vehicle, the type of goods being transported, temperature control requirements, and destination information. It then matches a backup vehicle that meets the temperature control requirements or the nearest emergency cold storage facility from the available alternative transport resource pool, generates a relay transport or emergency warehousing scheduling plan, and pushes the plan to the relevant operator terminals to minimize the risk of cargo damage caused by equipment failure.

[0027] The user application layer and the cloud platform layer are connected through API interfaces, providing functions such as real-time data display across multiple terminals, abnormal alarm push notifications, historical data tracing, and compliance report generation, to meet the differentiated usage needs of different user roles (such as drivers, cold storage managers, logistics dispatchers, quality inspectors, and regulatory agencies).

[0028] Compared with existing technologies, this invention has the following advantages: This invention comprises a multi-source sensing layer, an edge computing gateway, a cloud platform layer, and a user application layer. Through distributed deployment of sensor arrays, edge preprocessing, and dynamic threshold adjustment, it achieves real-time acquisition, processing, and early warning of temperature data across the entire cold chain. This solves the problems of data fragmentation, monitoring blind spots, limited transmission, and weak intelligent early warning in the prior art. It has the advantages of achieving real-time, dynamic monitoring of temperature across the entire cold chain, effectively reducing data silos, and improving the accuracy and timeliness of temperature monitoring and early warning. It provides full-chain coverage, eliminating monitoring blind spots. The system covers the entire cold chain process of warehousing, transportation, and distribution, achieving online temperature monitoring from goods entering the warehouse to final delivery, solving the problem of data silos at each stage. The multi-source sensing data is highly reliable. A three-dimensional monitoring network of spatial temperature field is formed using a multi-point temperature sensor array. Combined with Grubbs' criterion verification and fusion processing by the edge computing gateway, abnormal sensor data is effectively eliminated, improving the accuracy and reliability of temperature data. The Edge Computing Gateway performs local data preprocessing, anomaly detection, and rapid response, controlling data response time to the millisecond level; the rotating door compression algorithm significantly reduces data upload volume; the cloud handles global analysis and long-term data management, achieving rational allocation of computing resources; it adopts a multi-level early warning strategy of "static threshold + dynamic change rate + trend prediction," with an adaptive threshold algorithm dynamically adjusting alarm thresholds based on changes in the external environment, and a lightweight AI model enabling 48-hour advance warning of equipment failures, transforming "post-event traceability" into "pre-event early warning"; it achieves real-time 3D visualization monitoring of the cold chain environment through digital twin technology, visually highlighting abnormal areas and generating disposal suggestions to improve management efficiency; blockchain evidence storage technology ensures the immutability and traceability of temperature data throughout the cold chain process, and the barcode verification function enhances consumer and regulatory trust, meeting compliance requirements such as GSP / GMP; in extreme cases such as refrigeration equipment failure, the system automatically generates relay transportation or emergency warehousing scheduling plans, effectively reducing the risk of cargo damage. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the overall architecture of the online temperature monitoring system for the entire cold chain of this invention.

[0030] Figure 2 This is a schematic diagram of the deployment of a multi-point temperature sensor array in the multi-source sensing layer of this invention.

[0031] Figure 3 This is a block diagram of the functional modules of the edge computing gateway in this invention.

[0032] Figure 4 This is a block diagram of the functional modules of the cloud platform layer in this invention.

[0033] Figure 5 This is a flowchart illustrating the multi-level early warning strategy in this invention.

[0034] Figure 6 This is a flowchart of the sensor data verification and fusion method in this invention.

[0035] Figure 7 This is a flowchart illustrating the adaptive threshold algorithm for dynamically adjusting alarm thresholds in this invention.

[0036] Figure 8 This is a schematic diagram of the visual interface of the digital twin module in one embodiment of the present invention. Detailed Implementation

[0037] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0038] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0039] Please see Figure 1 As shown, this application proposes an Internet of Things-based online temperature monitoring system for the entire cold chain, comprising: The multi-source sensing layer is deployed at cold chain storage nodes, cold chain transportation carriers, and terminal delivery nodes. This multi-source sensing layer includes a multi-point temperature sensor array, position sensors, and environmental sensors. The multi-point temperature sensor array is deployed in a distributed manner at different depths and orientations in the space, forming a three-dimensional monitoring network of the spatial temperature field. The multi-source sensing layer sends the collected temperature data to the edge computing gateway through a LoRa wireless communication module and to the mobile terminal through a BLE communication module. An edge computing gateway, deployed on cold chain transport vehicles or cold chain storage nodes, receives temperature data sent by the multi-source sensing layer. The edge computing gateway includes a data preprocessing module, an anomaly detection module, and a local decision-making module. The data preprocessing module performs noise reduction, outlier removal, and data compression on the received raw temperature data. The anomaly detection module incorporates an adaptive threshold algorithm. This algorithm presets a basic temperature safety range based on the type of goods being transported and the current cold chain stage, and acquires external temperature data collected by environmental sensors in real time. It calculates the rate of change of external temperature, and when the rate of change exceeds a preset threshold, it automatically narrows the upper and lower limits of the safety range proportionally to dynamically adjust the temperature alarm threshold. The local decision-making module immediately triggers a local alarm and automatically adjusts the refrigeration equipment when a temperature anomaly is detected. The cloud platform layer interacts with the edge computing gateway through a 4G / 5G communication module to receive temperature data that has been preprocessed at the edge. The cloud platform layer includes a time-series database engine, an intelligent analysis module, a digital twin module, and a blockchain evidence storage module. The user application layer connects to the cloud platform layer via API interfaces, providing functions such as real-time data display across multiple terminals, abnormal alarm push notifications, historical data tracing, and compliance report generation.

[0040] For ease of understanding, the following explains some key terms in this embodiment: The multi-source sensing layer is configured as the bottom layer of the cold chain temperature online monitoring system. Its function is to be deployed at cold chain storage nodes, cold chain transportation carriers, and terminal delivery nodes, responsible for collecting various physical data in the cold chain environment in real time. This layer achieves sensing of the cold chain environment by integrating multiple sensors.

[0041] The multi-point temperature sensor array is designed as a distributed sensor network, with its sensor nodes deployed at different depths and orientations within a specific space (such as a refrigerated truck compartment or cold storage). This deployment method enables the array to achieve three-dimensional monitoring of the spatial temperature field, thereby acquiring temperature distribution data and avoiding blind spots that may exist in single-point monitoring.

[0042] Edge computing gateways are configured to be deployed on cold chain transport vehicles or cold chain storage nodes, serving as a connection between the multi-source perception layer and the cloud platform layer. This gateway performs computational tasks such as data preprocessing, anomaly detection, and local decision-making near the data source, aiming to reduce cloud load, lower network bandwidth requirements, and shorten response time.

[0043] An adaptive threshold algorithm is built into the anomaly detection module of the edge computing gateway. Its function is to adjust the upper and lower limits of temperature alarms based on changes in the cold chain process, cargo type, and external environment. This algorithm enables the system to adapt to various cold chain environmental conditions, improving the accuracy of early warnings.

[0044] Please see Figure 4 As shown, the cloud platform layer is configured as the core processing and management center of the system. It receives pre-processed temperature data through data interaction with the edge computing gateway. This layer provides services such as data storage, analysis, digital twin modeling, and blockchain notarization, supporting the mining and application of cold chain data.

[0045] Please see Figure 8 As shown, the digital twin module is configured to build a 3D digital model of the physical entity of the cold chain in the cloud and map the temperature field distribution in the physical world in real time. Through this module, users can view the real-time status of the cold chain environment and assist in anomaly location and decision-making.

[0046] The blockchain-based evidence storage module is configured to perform hash calculations on key temperature data throughout the cold chain process and store it on the blockchain. This module leverages the immutability of blockchain technology to ensure the authenticity, integrity, and traceability of cold chain data, thereby enhancing data credibility.

[0047] The user application layer is configured as the upper layer of the system and connects to the cloud platform layer through API interfaces. This layer provides users with functions such as real-time data display across multiple terminals, anomaly alarm push notifications, historical data tracing, and compliance report generation, serving as the interface for user interaction with the system.

[0048] The main technical features of this embodiment are described in detail below: Multi-source sensing layers are deployed at various key nodes in cold chain logistics. For example, in large cold storage facilities, sensor nodes can be fixedly installed between shelves or on walls; in refrigerated transport vehicles, sensor nodes can be attached or suspended in different locations inside the vehicle compartment; and in the last-mile delivery stage, sensor nodes can be integrated into delivery boxes or handheld devices.

[0049] This layer can include various types of sensors. For example, a multi-point temperature sensor array could consist of multiple thermistors or thermocouples, which are manually placed in different areas of the refrigerated space to obtain temperature readings at multiple points. A location sensor could be a GPS module used to report the device's geographic coordinates. An environmental sensor could be a stand-alone humidity sensor used to monitor air humidity.

[0050] The deployment of a multi-point temperature sensor array can be achieved through manual planning, with sensor nodes evenly distributed at the top, middle, and bottom of the cold storage space, as well as in different directions such as front, back, left, and right, thus forming a temperature monitoring grid covering the entire space.

[0051] The LoRa wireless communication module can use the standard LoRaWAN protocol to periodically send temperature data collected by the sensor to the edge computing gateway. For example, the sensor collects temperature data every 5 minutes and sends the data packets out through the LoRa module.

[0052] BLE communication modules can be used for data transmission with mobile terminals at close range. For example, when an inspector approaches a refrigerated truck compartment with a handheld mobile terminal, the BLE module can automatically synchronize the current or cached temperature data to the mobile terminal.

[0053] An edge computing gateway can be an embedded device installed in the control box of a refrigerated truck compartment or in the computer room of a cold storage facility. This gateway continuously listens for and receives temperature data packets sent from the multi-source sensing layer via a LoRa module.

[0054] The data preprocessing module can perform preliminary processing on the received raw temperature data. For example, it can denoise the data using a simple moving average filter to eliminate random noise; it can directly remove data points that exceed the physically reasonable range (such as -100℃ or +100℃) by setting a fixed threshold; and it can compress the data by simply reducing the sampling rate or using differential coding to reduce the amount of data.

[0055] The anomaly detection module can incorporate an adaptive threshold algorithm. This algorithm can pre-store a table mapping cargo types to baseline temperature safety ranges; for example, for frozen meat, the baseline range is -18℃ to -22℃. The system can select the corresponding baseline range based on the cargo type manually input by the user and the current cold chain stage (e.g., during transportation). Simultaneously, the algorithm can periodically acquire external temperature data from environmental sensors and calculate its average rate of change over the past hour. When this rate of change exceeds a fixed preset value (e.g., 2℃ per hour), the algorithm can simply shrink the upper and lower limits of the baseline safety range by 0.5℃ each, thus achieving dynamic threshold adjustment.

[0056] When the anomaly detection module determines that the temperature data exceeds the currently dynamically adjusted alarm threshold, the local decision module can immediately trigger a local audible and visual alarm and send instructions directly to the controller of the refrigeration equipment via relay control signals, such as lowering the set temperature of the refrigeration equipment by 1°C or increasing the fan speed.

[0057] The cloud platform layer can be deployed on a remote server cluster and connect to the edge computing gateway via a 4G / 5G communication module. The edge computing gateway periodically uploads pre-processed temperature data to the cloud platform.

[0058] A time-series database engine can be a database specifically designed for storing time-series data, such as InfluxDB or TimescaleDB. It is responsible for storing and managing the massive amounts of temperature data received from the edge computing gateway and supports time-range queries.

[0059] The intelligent analysis module can analyze historical temperature data stored in a time-series database. For example, it can predict temperature trends for the next few hours using a simple linear regression model, or identify different levels of temperature anomalies by setting multiple fixed thresholds.

[0060] The digital twin module can construct a simplified two-dimensional floor plan of a refrigerated truck or cold storage facility in the cloud. When temperature data is received, the current temperature value can be displayed on the floor plan at the corresponding sensor locations.

[0061] The blockchain evidence storage module can perform hash calculations on the received key temperature data and record the hash value in a centralized log system to provide basic audit trail capabilities.

[0062] The user application layer can be a mobile application or a web application, which interacts with the cloud platform layer through standard RESTful API interfaces.

[0063] This application provides a real-time temperature data display interface, such as displaying the current temperature of each cold chain node in a list format. When the cloud platform detects an anomaly, it can push alarm notifications to users via SMS or email. Users can query historical temperature data for a specific time period through the application and generate reports containing temperature records.

[0064] The IoT-based online temperature monitoring system for the entire cold chain proposed in this application achieves coverage of warehousing, transportation, and distribution by deploying a multi-source sensing layer throughout the cold chain process, solving the problems of fragmented monitoring and data silos in traditional systems. The introduction of an edge computing gateway enables data preprocessing and anomaly detection near the source, reducing network bandwidth load and response latency. In particular, the built-in adaptive threshold algorithm dynamically adjusts alarm thresholds based on cargo type and external environment, improving the accuracy of early warnings. Combined with cloud platform-level data analysis and data storage capabilities, and user application-level interaction, this system provides end-to-end temperature monitoring, early warning, and data traceability, ensuring the quality and safety of cold chain goods.

[0065] Please see Figure 2As shown, this application further proposes an IoT-based online temperature monitoring system for the entire cold chain. The multi-point temperature sensor array includes at least three temperature sensor nodes, strategically deployed at different depths and orientations within the space of cold chain storage nodes, cold chain transport vehicles, or terminal delivery nodes. For example, in a refrigerated truck compartment, sensor nodes can be placed at the top, middle, and bottom, as well as the front, middle, and rear, forming a three-dimensional monitoring network capable of capturing spatial temperature gradients and local hot or cold spots. To ensure data accuracy and reliability, the measurement accuracy of each temperature sensor node is set to no less than ±0.5℃, and the resolution to no less than 0.1℃. This allows the system to accurately sense minute temperature fluctuations, meeting the stringent temperature control requirements for high-value or highly sensitive cold chain goods. Simultaneously, the sampling frequency can be configured from 1Hz to 1 time / minute, allowing the system to flexibly adjust the data collection density according to actual needs (such as cargo sensitivity and cold chain links), optimizing resource consumption while ensuring data real-time performance.

[0066] The edge computing gateway also includes a multi-protocol compatibility module and a sensor data verification module. The multi-protocol compatibility module is a crucial component of the edge computing gateway, primarily enhancing its interoperability and scalability. By supporting multiple communication protocols, this module enables the edge computing gateway to interact not only with data transmitted from LoRa and BLE wireless communication modules, but also with other temperature control devices, refrigeration units, or auxiliary sensors (e.g., devices using Modbus, CAN bus, Ethernet, or other industry-standard protocols) that may exist in the cold chain environment. This provides flexibility for integrating devices from different manufacturers and reserves interfaces for future technology upgrades and functional expansions, avoiding system isolation problems caused by protocol incompatibility.

[0067] Please see Figure 6As shown, the sensor data verification module is a key component in the edge computing gateway used to improve the reliability of temperature data. Its core function is to perform quality checks and fusion processing on multiple temperature data points from a multi-point temperature sensor array to eliminate outliers and generate more representative fused temperature values. Specifically, after receiving temperature values ​​from various sensors collected simultaneously, the module first sorts them in ascending order and calculates the mean and standard deviation of the dataset. These statistics provide the basis for subsequent outlier detection. Next, based on a preset confidence probability (e.g., 95% or 99%) and the number of valid sensors currently participating in the detection, the module obtains the critical value of the Grubbs criterion by consulting a pre-stored statistical table. This critical value is the statistical basis for determining whether data is an outlier. Subsequently, the module calculates the Grubbs statistics for the maximum and minimum values ​​in the sorted data. The Grubbs statistic measures the degree to which extreme values ​​deviate from the mean. By comparing the calculated statistics with the previously obtained critical value, if the statistics exceed the critical value, the corresponding maximum or minimum value is determined to be an outlier and removed from the dataset. To ensure outliers are completely removed, the above steps are repeated until no data in the remaining dataset is identified as an outlier. This iterative processing mechanism effectively handles situations where multiple outliers exist in the dataset. Finally, the arithmetic mean of all retained valid temperature data is calculated, and this mean is used as the merged temperature value at that moment. This merged temperature value eliminates the interference of outliers and more accurately reflects the true temperature conditions of the cold chain environment.

[0068] Through the above technical solution, this application effectively solves the problems of potential errors and outliers in distributed multi-point temperature sensor data, significantly improving the accuracy and reliability of cold chain temperature monitoring. The multi-point temperature sensor array, with its high precision, high resolution, and configurable sampling frequency, provides detailed temperature field data within the cold chain space, laying the foundation for a comprehensive understanding of temperature control conditions. Furthermore, the newly added multi-protocol compatible module in the edge computing gateway enhances the system's openness and integration capabilities, enabling flexible access to more types of cold chain equipment. More importantly, the sensor data verification module, by introducing the Grubbs criterion for iterative outlier detection and removal, can identify and eliminate erroneous readings caused by sensor failure, local interference, or transient fluctuations from multiple sensor data. This rigorous data verification process ensures that the subsequent temperature data fusion results are highly reliable and representative. The final fused temperature value eliminates interference from abnormal data, providing high-quality input for the anomaly detection module and local decision-making module of the edge computing gateway. This avoids false alarms or missed alarms caused by erroneous data, enabling the system to more accurately determine whether the cold chain environment is within a safe range and take timely and effective local intervention measures, such as automatically adjusting refrigeration equipment, thereby ensuring the quality and safety of cold chain goods.

[0069] Please see Figure 3 As shown, this application further proposes that the edge computing gateway has a built-in lightweight AI model. This model takes the preprocessed temperature sequence and the operating parameters of the chiller unit as input, uses a temporal convolutional network or a long short-term memory network to extract the temporal features of temperature changes, and combines historical fault label data to train an anomaly detection and fault classifier. By analyzing the temperature change trend and the historical data of the sensors, it identifies temperature anomaly patterns and realizes the prediction of potential faults of the chiller unit, with an early warning time of no less than 48 hours.

[0070] This lightweight AI model refers to an optimized artificial intelligence model with lower computational resource requirements and memory footprint. Its design goal is to run efficiently on the limited hardware resources of edge computing gateways to achieve rapid local inference and decision-making. The model can be optimized using techniques such as model pruning, quantization, and knowledge distillation to significantly reduce model complexity and operational overhead while maintaining predictive accuracy. The preprocessed temperature series refers to high-quality and representative time-series temperature data that has undergone noise reduction, outlier removal, and data compression by the data preprocessing module. Refrigeration unit operating parameters include, but are not limited to, key indicators reflecting the operating status of refrigeration equipment, such as compressor start / stop status, fan speed, refrigerant pressure, current, and voltage. Using this data as input to the AI ​​model provides comprehensive contextual information, enabling it not only to perceive temperature changes but also to understand the internal causes of these changes, thereby more accurately identifying potential failure modes.

[0071] Temporal Convolutional Networks (TCNs) are deep learning models specifically designed for processing time-series data. They capture long-term dependencies and local patterns in time series data through dilated causal convolutional layers, offering advantages such as strong parallel computing capabilities and controllable receptive fields. Long Short-Term Memory Networks (LSTMs) are a variant of Recurrent Neural Networks (RNNs). By introducing gating mechanisms (input gate, forget gate, output gate), they effectively address the gradient vanishing or exploding problems of traditional RNNs when processing long sequences, enabling them to learn and remember long-term dependencies in time series data. Both network structures excel at extracting complex temporal features from continuous temperature sequences and equipment operating parameters, such as periodic temperature fluctuations, trend changes, abrupt change patterns, and correlations with equipment operating status. This process involves a supervised learning paradigm, which trains the AI ​​model using previously accumulated datasets labeled as normal or specific fault types. Historical fault label data includes what faults (e.g., refrigerant leaks, compressor failures, fan failures, etc.) occurred in the refrigeration unit under different operating conditions, along with the corresponding temperature sequences and operating parameter data. Through training, the model learns the differences between normal operating modes and various fault modes, thus building a classifier capable of distinguishing between normal states, potential abnormal states, and specific fault types. The AI ​​model analyzes the input real-time preprocessed temperature sequence and refrigeration unit operating parameters using its internally learned pattern recognition capabilities. This includes, but is not limited to, monitoring the rate of temperature rise / fall, fluctuation amplitude, deviation from the setpoint, and the correlation between these changes and operating parameters such as refrigeration unit start / stop and load changes. By comparing the data with historical normal and fault patterns learned during training, the model can identify whether there are abnormal patterns in the current data stream that match known fault patterns. For example, a sustained slow rise in temperature may indicate a decrease in refrigeration efficiency, while drastic fluctuations may point to sensor malfunction or refrigerant circulation problems.

[0072] Through the aforementioned technical solution, the edge computing gateway is no longer limited to passively responding to real-time temperature anomalies, but can proactively predict potential failures of refrigeration units. This lightweight AI model utilizes pre-processed high-quality temperature sequences and refrigeration unit operating parameters, deeply mining the temporal features in the data through temporal convolutional networks or long short-term memory networks, and combining this with historical fault label data for training, thereby constructing a powerful anomaly detection and fault classification capability. When the model identifies temperature change trends and anomaly patterns related to potential faults, the system can issue early warnings, providing users with at least 48 hours of warning time. This enables cold chain managers to take preventative maintenance measures before failures occur, such as arranging equipment maintenance, replacing worn parts, or adjusting refrigeration strategies, thereby effectively avoiding cold chain interruptions and cargo damage caused by sudden equipment failures, significantly improving the reliability and safety of cold chain transportation and warehousing, and reducing operational risks and economic losses.

[0073] Please see Figure 4 As shown, this application proposes an Internet of Things-based online temperature monitoring system for the entire cold chain, whose cloud platform layer further includes an intelligent analysis module, which is further refined into a temperature trend prediction unit and a multi-level early warning unit.

[0074] Specifically, the intelligent analysis module is a core component of the cloud platform layer. Its main function is to perform in-depth mining and advanced analysis of the massive amounts of cold chain temperature data received. This module aims to go beyond simple threshold judgments, using sophisticated algorithms and models to extract valuable information from the data to support smarter decision-making and more refined management. Its implementation typically involves a distributed computing framework deployed on cloud servers, capable of processing large-scale time-series data and providing scalable analytical capabilities.

[0075] The temperature trend prediction unit is a functional component of the intelligent analysis module. Its function is to predict temperature changes over a future period using historical data and external information. This enables the system to shift from passive response to proactive early warning. Based on historical temperature sequences in the time-series database engine and accessed external weather forecast data, this unit uses either the ARIMA or Prophet model to predict temperature change trends over a preset time period. The ARIMA model is a classic statistical time-series prediction model suitable for data with trends and seasonal patterns. It achieves stationarity by differencing the time series and then combines autoregressive and moving average components for modeling. The Prophet model, developed by Facebook, is a time-series prediction model particularly suitable for business time-series data with significant seasonal and holiday effects. It uses a decomposable model for prediction and exhibits good robustness to missing and outlier values. During the prediction process, the system automatically performs model parameter order determination and seasonal decomposition, i.e., determining the optimal parameter values ​​for each component of the model and identifying and removing seasonal cycles to improve prediction accuracy. External meteorological forecast data, such as ambient temperature, humidity, wind speed, and solar radiation intensity, can be input into the prediction model as external covariates, which can significantly improve the accuracy of temperature prediction.

[0076] Please see Figure 5As shown, the multi-level early warning unit is another important component of the intelligent analysis module. It is responsible for triggering different levels of alarms based on different temperature anomalies and prediction results. This hierarchical early warning mechanism provides more refined risk management, enabling operators to take corresponding countermeasures according to the risk level. The multi-level early warning unit adopts the following multi-level early warning strategy: When the real-time collected temperature data directly exceeds the preset basic temperature safety range, the system will trigger a level one alarm, indicating that the current temperature has deviated from the safe range and requires immediate attention. When the rate of change of temperature data per unit time exceeds the dynamic change rate threshold set according to the thermal sensitivity characteristics of the goods, the system will trigger a level two alarm. This alarm focuses on the trend and speed of temperature change. Even if the temperature has not completely exceeded the safe range, its rapid change may indicate potential risks, especially for goods sensitive to the rate of temperature change. Furthermore, when the temperature trend prediction unit predicts that the temperature will exceed the safe range in the future, the system will trigger a level three alarm, which is the highest level of alarm. Based on the prediction results, it provides an advanced warning, buying valuable time for taking preventive measures.

[0077] Through the aforementioned technical solutions, this system upgrades from traditional passive temperature monitoring and post-event alarms to proactive risk prediction and tiered early warning. The temperature trend prediction unit utilizes historical temperature sequences and external meteorological forecast data, employing advanced time-series models to accurately predict future temperature changes, enabling the system to identify potential temperature exceedance risks in advance. Based on this, a multi-level early warning unit combines real-time temperature, temperature change rate, and prediction results to construct a hierarchical early warning mechanism. When temperature data directly exceeds the safe range, a Level 1 alarm is triggered for immediate response; when the temperature change rate is too rapid, a Level 2 warning is triggered to indicate potential risks; and when a future temperature exceedance of the safe range is predicted, a Level 3 warning is triggered, providing operators with sufficient lead time to take preventative intervention measures, such as adjusting transportation routes, increasing refrigerant, or arranging advance cargo transfer. This proactive early warning and tiered response mechanism significantly enhances the risk management capabilities of cold chain logistics, effectively avoids cargo losses due to temperature anomalies, ensures the quality and safety of cold chain goods, and optimizes the efficiency and cost of emergency response.

[0078] Please see Figure 8As shown, this application further proposes that the digital twin module in the cloud platform layer constructs a three-dimensional digital model of the cold chain transportation carrier or storage space in the cloud, maps the temperature field distribution collected and fused by the multi-point temperature sensor array in real time, and displays the temperature distribution status in the three-dimensional digital model in the form of a heat map; when the edge computing gateway or the intelligent analysis module detects a local temperature anomaly, the digital twin module marks the abnormal area in the three-dimensional model with visual highlighting, and generates handling suggestions including adjusting the refrigeration setpoint, checking the integrity of the door seal, and guiding the nearest docking inspection according to the preset expert rule base.

[0079] Specifically, the digital twin module achieves bidirectional mapping and interaction between the physical and digital worlds by accurately modeling and simulating physical entities (such as refrigerated trucks and cold storage facilities) in digital space. When constructing a 3D digital model of a cold chain transport vehicle or storage space, methods such as CAD model import, 3D scanning, or manual modeling can be used to create a 3D geometric model containing key information such as internal structure, shelf layout, and refrigeration equipment location. This model not only carries geometric information but also includes physical attributes and logical relationships. To map the temperature field distribution collected and fused by a multi-point temperature sensor array in real time, the system combines multiple temperature data collected by the multi-point temperature sensor array and uses interpolation algorithms (such as Kriging interpolation, inverse distance weighted interpolation, or radial basis function interpolation) or finite element analysis to extend the discrete temperature point data to the entire 3D space, forming a continuous temperature field. Simultaneously, low latency in data transmission and processing is ensured, allowing the temperature field in the digital model to reflect the true state of the physical world in a timely manner. Subsequently, the system displays the temperature distribution in the form of a heat map in the three-dimensional digital model. This involves binding the mapped temperature field data to the three-dimensional model and rendering a heat map on the surface or inside the three-dimensional model using color gradients (e.g., from blue to red to represent low to high temperature) or isotherms. This allows users to view the temperature distribution at different angles and depths by rotating and scaling the model, thus providing an intuitive and rapid presentation of the temperature gradient and distribution within the cold chain space.

[0080] When the edge computing gateway or the intelligent analysis module detects a localized temperature anomaly, the digital twin module receives the alarm and, based on the sensor location or coordinates of the abnormal area, applies visual effects such as color flashing, highlighting, and transparency changes to the corresponding area in the 3D model. This visual highlighting of the abnormal area allows for rapid location of the anomaly, improving troubleshooting efficiency. Furthermore, the system generates handling suggestions based on a pre-defined expert rule base, including adjusting the cooling setpoint, checking door seal integrity, and guiding nearby docking for inspection. The expert rule base pre-constructs a knowledge base containing various anomalies and their corresponding handling solutions. Rules can be based on "IF-THEN" logic, such as "IF Localized temperature continues to rise AND Door magnetic sensor indicates door is open THEN: Check door seal integrity." Once an anomaly is detected and located, the digital twin module queries the expert rule base based on the anomaly type, location, and duration to match the most suitable handling suggestion. These suggestions can include: remotely adjusting the set temperature of refrigeration equipment, prompting to check the seals of refrigerated truck doors or cold storage doors, and recommending the nearest repair shop or emergency cold storage for inspection based on GPS location information, providing operators with intelligent and actionable guidance for handling abnormalities.

[0081] Through the above technical solutions, a three-dimensional digital model of the cold chain transport vehicle or storage space is constructed in the cloud. This model is then used to map and fuse the temperature field distribution collected by a multi-point temperature sensor array in real time, visually displaying the temperature status in the form of a heat map. This allows operators to clearly and comprehensively understand the temperature distribution within the cold chain environment, overcoming the problem of traditional monitoring systems that only provide discrete data and lack spatial visualization. When the edge computing gateway or intelligent analysis module detects a local temperature anomaly, the digital twin module can immediately and accurately mark the abnormal area in the three-dimensional model using visual highlighting, greatly improving the efficiency and accuracy of anomaly location. Furthermore, based on a preset expert rule base, the system can intelligently generate specific handling suggestions, including adjusting the refrigeration setpoint, checking the integrity of door seals, and guiding nearby docking for inspection. This provides operators with timely and effective decision support, avoiding the lag and uncertainty of manual judgment, effectively reducing the risk of cargo damage, and improving the intelligence level and emergency response capabilities of cold chain management.

[0082] This application further proposes that the cloud platform layer includes a blockchain evidence storage module. The blockchain evidence storage module performs hash calculations on key temperature data throughout the cold chain process and stores it on the blockchain. The key temperature data includes temperature records, abnormal event records, and handover confirmation information for each stage. The blockchain evidence storage module calculates a hash value for each key data item, packages the hash value into a blockchain transaction, and returns an evidence storage identifier. A QR code is generated based on the evidence storage identifier, which can be scanned by consumers and regulatory agencies to verify the tamper-proof cold chain data evidence.

[0083] The blockchain evidence storage module is a system component based on distributed ledger technology. Its core function is to ensure the immutability, traceability, and transparency of data using cryptographic principles and consensus mechanisms. This module can be deployed at the cloud platform layer as part of the cloud platform layer, responsible for receiving and processing critical data from edge computing gateways. Its implementation can employ consortium blockchain or private blockchain technology to meet the data privacy and performance requirements of the cold chain industry, while ensuring joint maintenance and supervision of data by participating parties. The critical temperature data throughout the cold chain process refers to temperature monitoring data crucial to the quality and safety of goods at each stage of cold chain logistics, records of abnormal events detected by the system, and confirmation information of goods handover between different stages. This data forms the basis for evaluating cold chain service quality, tracing responsibility, and meeting regulatory requirements. Hash operations are the process of converting data of arbitrary length into a fixed-length unique string (hash value) using a hash algorithm. Even minor changes to the original data will result in a significant change in its hash value. Storing the hash values ​​of these critical data on the blockchain, rather than directly storing the raw data, effectively reduces the storage burden on the blockchain. Simultaneously, leveraging the blockchain's immutability, it ensures the integrity of the raw data is guaranteed once hashed and uploaded. When the cold chain system generates a piece of critical data, the blockchain notarization module immediately performs a hash operation, generating a unique hash value. Subsequently, this hash value, along with metadata such as timestamp and data type, is encapsulated into a blockchain transaction. This transaction is submitted to the blockchain network, verified through the consensus mechanism, packaged into a new block, and added to the chain. Once the transaction is successfully uploaded to the chain, the system generates a unique notarization identifier, such as the transaction hash or block height, serving as a unique credential for the data on the blockchain. This identifier can be used for subsequent data queries and verification. To facilitate quick and convenient verification of the authenticity and integrity of cold chain data by non-technical personnel, the system can automatically generate a QR code containing this identifier based on the notarization identifier returned after uploading. Consumers or regulatory agencies can simply scan the QR code using a smartphone or other device to access the blockchain network through the user application layer or a dedicated query interface to query the on-chain record corresponding to the notarization identifier. By comparing the hash value on the chain with the hash value of the original data, it is possible to verify whether the original data has been tampered with, thereby enabling public verification of the immutability of cold chain data.

[0084] This application further proposes that, in an IoT-based online temperature monitoring system for the entire cold chain, although the local decision-making module of the edge computing gateway can automatically adjust the refrigeration equipment to attempt to restore a safe temperature when an abnormal temperature is detected, in some extreme cases, such as when the refrigeration equipment malfunctions or the external environment has an excessive heat load, the local adjustment may not be able to effectively resolve the persistent temperature abnormality, which may lead to the risk of cold chain interruption and damage to goods.

[0085] To address the aforementioned issues, this application proposes an improved solution: the cloud platform layer further includes a collaborative scheduling module. This module is a core functional unit of the cloud platform layer, and its primary responsibility is to coordinate and allocate external resources during cold chain transportation or warehousing when severe temperature anomalies exceeding local processing capacity occur, ensuring the continuity of the cold chain and the safety of goods. It integrates decision-making logic, resource management, and communication interfaces, aiming to achieve rapid response and effective intervention in emergencies.

[0086] Specifically, the collaborative scheduling module is triggered when the system detects a persistent temperature anomaly in a cold chain transport vehicle and the local decision-making module is unable to restore the temperature to a safe range by adjusting the refrigeration equipment. This condition is the trigger mechanism for the collaborative scheduling module; it indicates that the system has identified a serious anomaly with potential risks that is not merely a momentary or minor temperature fluctuation, but one that the local decision-making module has been unable to resolve despite its intervention attempts. Such persistent anomalies typically mean that the local refrigeration equipment may be malfunctioning, its cooling capacity insufficient, or the external environmental conditions are too harsh, requiring a higher level of intervention.

[0087] Once triggered, the collaborative scheduling module immediately acquires the real-time location of the abnormal vehicle, the type of goods being transported, temperature control requirements, and destination information. To formulate an effective emergency response plan, the collaborative scheduling module needs a comprehensive understanding of the abnormal vehicle's key information. Real-time location information is used to determine the nearest emergency resource point; the type of goods being transported and temperature control requirements are key parameters for matching alternative resources (such as backup vehicles or emergency cold storage), ensuring that the alternative plan meets the specific temperature requirements of the goods; destination information helps plan the relay transportation route or subsequent processing after emergency storage.

[0088] Subsequently, the collaborative scheduling module will match suitable backup vehicles or nearby emergency cold storage facilities from the available alternative transportation resource pool. The alternative transportation resource pool is a pre-established database containing various types of backup cold chain transport vehicles and emergency cold storage facilities. Based on the specific needs of the abnormal vehicle (such as temperature control range, capacity, and geographical location), the collaborative scheduling module will intelligently search and filter within this resource pool to select the most suitable backup vehicle or the nearest emergency cold storage facility with corresponding temperature control capabilities. The matching process may involve route optimization and resource availability assessment based on a Geographic Information System (GIS).

[0089] Once a suitable alternative resource is matched, the collaborative scheduling module will generate a relay transportation or emergency warehousing scheduling plan. Relay transportation plans typically include specifying the departure time of the backup vehicle, the pick-up location, the transportation route, and the handover process with the original carrier; emergency warehousing plans will specify the nearest emergency cold storage facility, the warehousing process, and the unloading and storage arrangements for the goods. These plans aim to provide clear and actionable instructions for the rapid transfer of threatened goods. Finally, the collaborative scheduling module will push the plan to the relevant operator terminals. The generated scheduling plan needs to be promptly communicated to the implementing personnel. Relevant operator terminals can include mobile devices or computer terminals of drivers, warehouse managers, dispatch center personnel, or customer service representatives, ensuring that information is delivered quickly and accurately so that operators can take immediate action.

[0090] Through the above technical solution, in the cold chain transportation or warehousing process, when the local decision-making module of the edge computing gateway is unable to restore the temperature to a safe range by adjusting the refrigeration equipment and a persistent temperature anomaly occurs, the collaborative scheduling module of the cloud platform layer can proactively intervene, obtain key information about the abnormal carrier, and intelligently match alternative transportation resources or emergency cold storage, generating and pushing a relay transportation or emergency warehousing scheduling plan. This effectively solves the problem that traditional cold chain monitoring systems lack advanced emergency response mechanisms when facing local equipment failures or extreme environmental conditions. By introducing collaborative scheduling, the system can transform from passive alarm to proactive intervention, ensuring that even in the most unfavorable circumstances, measures can be taken quickly to minimize losses caused by temperature runaway, protect the integrity of the cold chain and the quality of goods, thereby significantly improving the reliability and resilience of cold chain logistics.

[0091] This application further proposes that the environmental sensor includes at least one of a humidity sensor, a door magnetic sensor, and a vibration sensor. The humidity sensor is used to monitor humidity data in the cold chain environment. During cold chain storage and transportation, humidity is one of the key factors affecting the quality of goods; excessively high or low humidity can lead to spoilage, condensation, or drying. Humidity sensors typically employ capacitive or resistive principles, outputting corresponding electrical signals by sensing changes in ambient humidity, thereby achieving real-time monitoring of the humidity in the cold chain environment. The door magnetic sensor is used to monitor the opening and closing status and duration of the door in the space it is located on. In the cold chain environment, frequent opening or prolonged periods without closing the door can lead to a significant loss of cold air, causing internal temperature fluctuations and thus affecting the quality of the goods. The door magnetic sensor typically consists of a magnet and a magnetic switch. When the door is closed, the magnet approaches the magnetic switch, causing it to close; when the door is open, the magnet moves away from the magnetic switch, causing it to open, thus achieving accurate sensing of the door's status. The vibration sensor is used to monitor impact vibration events exceeding a preset threshold during transportation. During cold chain transportation, severe vibration or impact can cause physical damage to goods and even affect the normal operation of refrigeration equipment. Vibration sensors typically employ piezoelectric effect or MEMS (microelectromechanical systems) technology to sense and quantify the acceleration or impact force experienced by an object during transportation. When the detected vibration intensity exceeds a preset safety threshold, the system can record the event.

[0092] Furthermore, the data preprocessing module of the edge computing gateway employs a rotating door compression algorithm for the raw sensor data. This algorithm approximates continuously collected time-series data points as a series of line segments by setting a compression deviation threshold. Specifically, the algorithm maintains a "rotating door." When a new data point enters, it checks whether it falls within the deviation threshold of the current line segment. If it is within the threshold, the data point is considered part of the current line segment and is not uploaded; if it exceeds the threshold, the current line segment ends, the new data point becomes the starting point of a new line segment, and the endpoint data of the previous line segment is uploaded. In this way, only the endpoint data of the line segment is uploaded, thereby achieving a compression rate of at least 80% for the raw data, reducing network bandwidth load.

[0093] Through the aforementioned technical solutions, this system achieves comprehensive perception of multi-dimensional parameters of the cold chain environment, not only temperature but also humidity, door status, and transportation impact. This allows for a more comprehensive assessment of the storage conditions and transportation safety of cold chain goods, and timely detection and early warning of potential non-temperature-related anomalies. Simultaneously, the edge computing gateway's data preprocessing module employs a rotating door compression algorithm, efficiently compressing this multi-source sensor data, significantly reducing data transmission volume, effectively alleviating network bandwidth load, reducing data transmission latency and costs, ensuring the real-time performance and reliability of critical data, and providing a high-quality and efficient data stream for deeper intelligent analysis at the cloud platform layer. This combination of multi-dimensional perception and efficient data transmission significantly improves the precision of cold chain monitoring and the economic efficiency of system operation.

[0094] This application further proposes that the LoRa wireless communication module adopts spread spectrum modulation technology to establish a wireless connection with the edge computing gateway in a metal enclosed car or cold storage environment, with a data packet loss rate of less than 0.3%; the BLE communication module is used to perform short-range data synchronization with a handheld mobile terminal during inspection or loading and unloading, realizing offline data retransmission and on-site parameter configuration.

[0095] Specifically, the LoRa wireless communication module employs spread spectrum modulation technology, which significantly enhances signal anti-interference capabilities and penetration by extending narrowband signals to a wider frequency band for transmission. For example, Chirp Spread Spectrum (CSS) modulation can be used, which encodes data using continuously varying frequency pulses, ensuring good transmission performance even in complex electromagnetic environments, such as enclosed metal car bodies or cold storage rooms with thick walls. This modulation method effectively overcomes the signal attenuation and multipath effects of traditional wireless communication in these shielded or reflective environments, ensuring the stability of the wireless connection between the multi-source sensing layer and the edge computing gateway. By optimizing antenna design and protocol parameters, the packet loss rate can be further controlled to below 0.3%, thereby guaranteeing the integrity and real-time performance of temperature data.

[0096] Meanwhile, the BLE communication module is configured to perform short-range data synchronization with a handheld mobile terminal during specific operations such as inspection or loading / unloading. BLE (Bluetooth Low Energy) is a low-power, short-range wireless communication technology ideal for data exchange at close range. In practical applications, when operators use a handheld mobile terminal (such as a smartphone or dedicated PDA) near cold chain equipment for inspection, the BLE communication module can quickly establish a connection, enabling offline data retransmission. This means that even if the LoRa network connection is unstable at certain times or to save bandwidth, sensor data can be stored locally and transmitted to the mobile terminal in one go via BLE, and then uploaded to the cloud platform layer by the mobile terminal. Furthermore, the BLE communication module supports on-site parameter configuration. Operators can directly adjust parameters such as sampling frequency and alarm thresholds of sensor nodes in the multi-source sensing layer through an application on their handheld mobile terminal, without needing remote operation through an edge computing gateway or cloud platform layer, greatly improving the flexibility and efficiency of on-site maintenance and management.

[0097] Through the aforementioned technical solutions, the LoRa wireless communication module employs spread spectrum modulation technology, significantly improving data transmission reliability in harsh radio frequency environments such as enclosed metal compartments or cold storage facilities. This effectively reduces data packet loss rates and ensures the continuity and accuracy of cold chain temperature data, thus providing a high-quality data source for real-time analysis and decision-making by the edge computing gateway. Simultaneously, the introduction of the BLE communication module provides the cold chain system with a flexible and convenient short-range interaction method, enabling efficient offline data retransmission and on-site parameter configuration in specific operational scenarios such as inspection and loading / unloading. This not only compensates for the shortcomings of long-distance communication in specific scenarios, improving the system's operational convenience and maintenance efficiency, but also, through the synergistic cooperation of the two communication methods, constructs a more robust, flexible, and adaptable online temperature monitoring system for the entire cold chain, ensuring temperature safety and compliance throughout the cold chain logistics process.

[0098] Please see Figure 7As shown, this application proposes an IoT-based online temperature monitoring system for the entire cold chain. The anomaly detection module of the edge computing gateway incorporates an adaptive threshold algorithm. This algorithm presets a basic temperature safety range based on the type of goods being transported and the current cold chain stage, and acquires external temperature data collected by environmental sensors in real time. It calculates the rate of change of external temperature, and when the rate of change exceeds a preset threshold, it automatically narrows the upper and lower limits of the safety range proportionally to dynamically adjust the temperature alarm threshold. However, in actual cold chain transportation or warehousing, the external ambient temperature may change drastically, causing rapid fluctuations in the internal temperature of the cold chain. If the alarm threshold is adjusted solely based on a simple judgment of the rate of change of external temperature, it may not reflect the actual risks faced by the goods in a timely and accurate manner. Especially for heat-sensitive goods, there may be a risk of delayed warnings or false alarms, affecting the reliability and efficiency of the cold chain.

[0099] To further optimize the performance of the aforementioned adaptive threshold algorithm, this application details the specific execution process of the algorithm when dynamically adjusting the temperature alarm threshold. Specifically, the anomaly detection module first determines whether the device is currently in the storage, transportation, or delivery stage using the location sensor and preset cold chain link logic. The location sensor can be a Global Positioning System (GPS) module, a BeiDou positioning module, or a Wi-Fi / base station-based positioning module to obtain the device's geographical location information. The preset cold chain link logic is a set of rules; for example, when the device is within a specific warehouse geofence, it is determined to be in the storage stage; when the device moves along a specific route, it is determined to be in the transportation stage; and when the device is within the terminal delivery area and frequently stops, it is determined to be in the delivery stage. This step aims to provide contextual information for subsequent temperature threshold adjustment, ensuring the accuracy and applicability of the threshold adjustment.

[0100] Based on this, the system reads the standard temperature safety lower and upper limits of the transported goods from the cargo information synchronized from the cloud platform layer. The cloud platform layer typically stores detailed attributes of various goods, including their temperature sensitivity, recommended storage and transportation temperature ranges, etc. After receiving cargo loading information, the edge computing gateway obtains the corresponding standard temperature safety range for the goods through data interaction with the cloud platform layer. This ensures that subsequent temperature monitoring and alarms are based on the characteristics of the goods themselves rather than general standards, improving the granularity of monitoring.

[0101] Simultaneously, the system receives real-time ambient temperature data collected by the environmental sensors and calculates the slope of the current ambient temperature change within a sliding time window. The environmental sensors may include temperature sensors, humidity sensors, etc., used to monitor the real-time temperature of the external environment of the cold chain carrier or the storage environment. The sliding time window refers to a continuous period of time, such as the past 5, 10, or 30 minutes, during which the system continuously collects ambient temperature data points. The slope of change can be calculated using linear regression, finite difference methods, or more complex time series analysis methods. For example, it can be achieved by calculating the slope of the least-squares fitted line of the temperature data points within the window, or simply by dividing the temperature difference between the beginning and end of the window by the length of the time window. This slope reflects the drasticness and direction of the ambient temperature change.

[0102] When the absolute value of the slope exceeds a preset slope threshold, the system proportionally narrows the upper and lower limits of the safety range based on the magnitude of the slope. The preset slope threshold is an empirical value or a critical value derived from historical data analysis, used to determine whether the external temperature change is sufficiently drastic to trigger a dynamic adjustment mechanism. The proportional coefficient is determined using a pre-calibrated cargo temperature rise sensitivity curve. This curve is obtained through experiments or simulations of the temperature response characteristics of different goods under varying external heat loads, describing the relationship between the rate of change of external temperature and the rate of temperature rise inside the cargo. For example, for highly sensitive goods, even a small external temperature change slope may require a larger proportional coefficient to quickly narrow the safety range. In this way, the alarm threshold can be narrowed in advance when the external heat load increases sharply, thus achieving early warning.

[0103] Through the above technical solution, the adaptive threshold algorithm of this application can achieve more refined and intelligent dynamic adjustment of temperature alarm thresholds. First, by judging the current cold chain link and reading the standard temperature safety range of the goods, the targetedness and accuracy of the alarm threshold setting are ensured. Second, by calculating the slope of the external temperature change within the sliding time window in real time, the system can accurately capture the drastic trend of external ambient temperature changes. More importantly, when the external temperature changes drastically, the system can determine a proportional coefficient based on the magnitude of the slope and a pre-calibrated temperature rise sensitivity curve of the goods, thereby proportionally narrowing the upper and lower limits of the safety range. This mechanism allows the alarm threshold to be dynamically and proactively adjusted according to the actual external heat load and the thermal sensitivity characteristics of the goods. For example, when the external ambient temperature rises sharply, even if the internal temperature of the cold chain has not yet reached the traditional alarm point, the system will narrow the safety range in advance due to the rapid increase in external heat load, thus triggering an early warning. This significantly improves the timeliness and accuracy of early warnings, avoiding the risk of cargo damage due to delayed warnings. It provides more reliable protection, especially for goods sensitive to temperature changes, thereby effectively reducing cold chain losses and improving the safety and efficiency of cold chain transportation.

[0104] The following example will provide a more detailed explanation of the above technical solution: Example 1: Cold chain transportation scenario for pharmaceuticals and vaccines A vaccine manufacturer deployed the system described in this invention during the cold chain transportation of vaccines. The storage temperature of vaccines must be strictly controlled between 2°C and 8°C; any temperature fluctuations outside this range may lead to a decrease in vaccine efficacy or even render the vaccine ineffective.

[0105] Sensing Layer Deployment: Six temperature sensor nodes are deployed in a distributed manner inside the interior space of a refrigerated truck compartment, located at the upper front, lower front, upper middle, lower middle, upper rear, and lower rear, forming a three-dimensional temperature sensing network covering three areas (front, middle, and rear) and two heights (vertical and horizontal). Simultaneously, door magnetic sensors are installed inside the compartment to monitor door opening and closing status, vibration sensors to detect abnormal impacts during transport, and humidity sensors to monitor compartment humidity. Each sensor node connects to an edge computing gateway installed in the driver's cab via a LoRa wireless communication module. LoRa spread spectrum modulation technology achieved stable data transmission within a 200-meter range inside the metal compartment, with an actual data packet loss rate of only 0.25%, far superior to traditional WiFi solutions.

[0106] Edge computing processing: The data preprocessing module of the edge computing gateway performs rotary gate compression on the received 6-channel temperature data. With the deviation threshold set at 0.1℃, the data compression rate reaches 85%, which significantly reduces the amount of data uploaded by 5G.

[0107] The sensor data verification module performs verification and fusion processing on the six temperature data points at each sampling time: First, the six temperature values ​​are arranged from smallest to largest, and the mean and standard deviation are calculated; a confidence probability of 95% is selected, and Grubbs' critical value table is consulted to obtain G(6,0.05)=1.887; the Grubbs' statistics for the maximum and minimum values ​​are calculated, and if the statistics exceed 1.887, the corresponding data is discarded; the arithmetic mean of the retained valid data is calculated as the fused temperature value. This mechanism successfully eliminates abnormal readings caused by individual sensors being in close contact with the carriage wall.

[0108] The adaptive threshold algorithm of the anomaly detection module presets a basic safety range of 2°C to 8°C based on the vaccine cargo information. When a vehicle drives from a temperature-controlled warehouse into a high-temperature summer environment, the external ambient temperature sensor detects that the external temperature rises sharply from 25°C to 38°C within 10 minutes. The slope of the change exceeds the preset slope threshold. The adaptive threshold algorithm automatically narrows the alarm upper limit from 8°C to 7.2°C, achieving early warning.

[0109] The lightweight AI model built into the edge computing gateway employs a long short-term memory (LSTM) network structure. The input layer receives fused temperature sequences and refrigeration unit operating parameters, extracts temporal features through three LSTM layers, and outputs fault classification results through a fully connected layer. Before deployment, the model is trained using labeled historical fault data, including typical fault modes such as refrigerant leakage, compressor wear, and condenser blockage. In one transportation mission, the AI ​​model analyzed the slowdown in temperature recovery rate and abnormal fluctuations in the compressor current spectrum, providing a 52-hour advance warning of an impending compressor bearing failure. This allowed maintenance personnel to replace the vehicle with a spare before departure, preventing a potential vaccine failure incident.

[0110] When the data from any temperature sensor exceeds the safe range after verification, the local decision module triggers an audible and visual alarm in the cab within 180ms and automatically adjusts the power of the refrigeration equipment.

[0111] Cloud processing and application: Temperature data preprocessed at the edge is uploaded to the cloud platform layer via a 5G network. The digital twin module of the cloud platform layer maps the three-dimensional temperature field distribution of the carriage in real time. When an anomaly is detected and a level one alarm is triggered, the corresponding area in the three-dimensional model is highlighted in red and flashing, and a handling suggestion window pops up, prompting the driver to check the cooling air outlet in that area.

[0112] The temperature trend prediction unit of the intelligent analysis module uses the Prophet model, taking into account historical temperature data from the past hour and path temperature data from external weather forecasts, to predict the temperature change trend for the next 15 minutes. During one transport operation, the system predicted that the temperature would exceed the upper limit of the safe zone in 8 minutes, triggering a level-three warning in advance. The driver, following the warning, increased the cooling capacity in advance, and the temperature stabilized and returned to the safe zone 6 minutes later, successfully averting a temperature exceedance event.

[0113] The blockchain evidence storage module calculates SHA-256 hash values ​​for temperature records, abnormal event records, and handover confirmation information every 5 minutes throughout the transportation process, and packages these hash values ​​into blockchain transactions. The consignee and regulatory agencies can verify the integrity and immutability of the temperature data on the blockchain by scanning the QR code on the waybill, thus meeting GSP compliance requirements.

[0114] Example 2: Multi-temperature zone cold chain monitoring scenario for fresh food A fresh food e-commerce platform has deployed the system described in this invention in its cold chain warehousing and distribution processes. The platform simultaneously sells three categories of goods in different temperature zones: frozen products (requiring temperatures below -18°C), refrigerated products (requiring temperatures between 0°C and 4°C), and fresh products (requiring temperatures between 8°C and 12°C).

[0115] Cold storage deployment: Deploy multi-point temperature sensor arrays within the cold storage area according to shelf zones, with 4 temperature sensor nodes set up for each zone, distributed at the four corners and center of the shelves. Install door magnetic sensors at each entrance and exit of the cold storage.

[0116] Deployment at the delivery end: In the refrigerated box of the delivery vehicle, according to the division of three temperature zones (frozen, refrigerated, and fresh), three temperature sensor nodes are deployed in each temperature zone, and the temperature zones are separated by heat-insulating partitions.

[0117] System operation: The cloud platform layer automatically configures differentiated temperature monitoring strategies based on the type of goods: the safe range for frozen products is -22℃ to -18℃; the safe range for refrigerated products is 0℃ to 4℃; and the safe range for fresh products is 8℃ to 12℃.

[0118] The edge computing gateway processes temperature data from three temperature zones simultaneously. The data preprocessing module performs rotating door compression and verification fusion on the sensor data of each temperature zone. The data of each temperature zone is processed independently without interference.

[0119] During system operation, the cloud platform's collaborative scheduling module successfully handled a refrigeration equipment failure. A refrigeration unit in the refrigerated zone of a delivery truck malfunctioned, causing the temperature in that zone to rise continuously. The local decision-making module determined that the temperature could not be restored to below 4°C through adjustment. The collaborative scheduling module obtained the vehicle's real-time location, the type of cargo being transported (fresh milk, refrigerated, 0-4°C), and the delivery destination information. The system matched an empty refrigerated truck 3 kilometers away that had already completed its delivery and met the temperature control requirements of -20°C to 10°C. Simultaneously, it matched a partner cold storage facility 5 kilometers away that could provide emergency storage. The system automatically generated a relay transport plan and an emergency storage plan, pushing them to the dispatcher. The dispatcher selected to dispatch the empty refrigerated truck for relay transport. The fresh milk was transferred to the backup vehicle before the temperature rose to 5.5°C, successfully preventing cargo damage.

[0120] By deploying the system of this invention, the fresh food e-commerce platform has achieved visualized temperature monitoring of the entire cold chain, shortened the response time for abnormal temperature events from an average of 30 minutes to less than 3 minutes, and reduced the damage rate by 67% compared to before deployment.

[0121] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the technical concept and spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1. An Internet of Things-based online temperature monitoring system for the entire cold chain, characterized in that, include: A multi-source sensing layer is deployed at cold chain storage nodes, cold chain transportation carriers, and terminal delivery nodes. The multi-source sensing layer includes a multi-point temperature sensor array, a position sensor, and an environmental sensor. The multi-point temperature sensor array is deployed in a distributed manner at different depths and orientations in the space, forming a three-dimensional monitoring network of the spatial temperature field. The multi-source sensing layer sends the collected temperature data to the edge computing gateway through a LoRa wireless communication module and to the mobile terminal through a BLE communication module. An edge computing gateway, deployed on a cold chain transport vehicle or cold chain storage node, receives temperature data sent by the multi-source sensing layer. The edge computing gateway includes a data preprocessing module, an anomaly detection module, and a local decision-making module. The data preprocessing module performs noise reduction, outlier removal, and data compression on the received raw temperature data. The anomaly detection module has a built-in adaptive threshold algorithm. This algorithm presets a basic temperature safety range based on the type of goods being transported and the current cold chain link, and acquires external temperature data collected by environmental sensors in real time to calculate the external temperature change rate. When the change rate exceeds a preset threshold, it automatically narrows the upper and lower limits of the safety range proportionally to dynamically adjust the temperature alarm threshold. The local decision module immediately triggers a local alarm and automatically adjusts the refrigeration equipment when it detects a temperature anomaly. The cloud platform layer interacts with the edge computing gateway through a 4G / 5G communication module to receive temperature data that has been preprocessed at the edge. The cloud platform layer includes a time-series database engine, an intelligent analysis module, a digital twin module, and a blockchain evidence storage module. The user application layer connects to the cloud platform layer via an API interface, providing functions such as real-time data display across multiple terminals, abnormal alarm push notifications, historical data tracing, and compliance report generation.

2. The IoT-based online temperature monitoring system for the entire cold chain according to claim 1, characterized in that, The multi-point temperature sensor array includes at least three temperature sensor nodes, deployed at different depths and orientations within the space. The temperature sensor nodes have a measurement accuracy of at least ±0.5℃, a resolution of at least 0.1℃, and a sampling frequency configurable from 1Hz to 1 time / minute. The edge computing gateway also includes a multi-protocol compatibility module and a sensor data verification module. After receiving multiple temperature data points from the multi-point temperature sensor array, the sensor data verification module performs the following verification and fusion steps: Arrange the temperature values ​​collected from each sensor at the same time in ascending order, and calculate the mean and standard deviation of the dataset; Based on the preset confidence probability and the current number of effective sensors, the critical value of Grubbs' criterion is obtained by looking up a table; Calculate the Grubbs statistic for the maximum and minimum values ​​in the sorted data, compare the statistic with the critical value, and if the value exceeds the critical value, the corresponding data is identified as an outlier and removed. Repeat the above steps until no data is identified as an outlier; The arithmetic mean of the retained effective temperature data is calculated and used as the fusion temperature value at that moment.

3. The IoT-based online temperature monitoring system for the entire cold chain according to claim 1, characterized in that, The edge computing gateway has a built-in lightweight AI model. This model takes the preprocessed temperature sequence and the operating parameters of the chiller unit as input, uses a temporal convolutional network or a long short-term memory network to extract the temporal features of temperature changes, and combines historical fault label data to train an anomaly detection and fault classifier. By analyzing the temperature change trend and historical sensor data, it identifies temperature anomaly patterns and realizes the prediction of potential faults of the chiller unit, with an early warning time of no less than 48 hours.

4. The IoT-based online temperature monitoring system for the entire cold chain according to claim 1, characterized in that, The intelligent analysis module of the cloud platform layer includes a temperature trend prediction unit and a multi-level early warning unit; The temperature trend prediction unit uses the historical temperature series in the time series database engine and the external meteorological forecast data to predict the temperature change trend in the future preset time period using the ARIMA model or the Prophet model. During the prediction process, the model parameters are automatically determined and seasonally decomposed. The multi-level early warning unit adopts the following multi-level early warning strategy: when the real-time collected temperature data exceeds the preset basic temperature safety range, a first-level alarm is triggered; when the rate of change of temperature data per unit time exceeds the dynamic change rate threshold set according to the thermal sensitivity characteristics of the goods, a second-level early warning is triggered. When the temperature trend prediction unit predicts that the temperature will exceed the safe range in the near future, it triggers a level three warning.

5. The IoT-based online temperature monitoring system for the entire cold chain according to claim 1, characterized in that, The digital twin module constructs a three-dimensional digital model of the cold chain transport vehicle or storage space in the cloud, maps the temperature field distribution collected and fused by the multi-point temperature sensor array in real time, and displays the temperature distribution status in the three-dimensional digital model in the form of a heat map. When the edge computing gateway or the intelligent analysis module detects a local temperature anomaly, the digital twin module marks the abnormal area in the three-dimensional model with visual highlighting, and generates handling suggestions based on a preset expert rule base, including adjusting the refrigeration setpoint, checking the integrity of the door seal, and guiding the vehicle to the nearest dock for inspection.

6. The IoT-based online temperature monitoring system for the entire cold chain according to claim 1, characterized in that, The blockchain evidence storage module performs hash calculations on key temperature data throughout the cold chain process and stores it on the blockchain. The key temperature data includes temperature records, abnormal event records, and handover confirmation information for each stage. The blockchain evidence storage module calculates a hash value for each key data item, packages the hash value into a blockchain transaction, and returns an evidence storage identifier. A QR code is generated based on the evidence storage identifier, which consumers and regulatory agencies can scan to verify the tamper-proof cold chain data evidence.

7. The IoT-based online temperature monitoring system for the entire cold chain according to claim 1, characterized in that, The cloud platform layer also includes a collaborative scheduling module. When a cold chain transport vehicle is detected to have a continuous temperature anomaly and the local decision-making module is unable to restore the temperature to a safe range by adjusting the refrigeration equipment, the collaborative scheduling module obtains the real-time location of the abnormal vehicle, the type of goods being transported, the temperature control requirements, and the destination information. It then matches a backup vehicle that meets the temperature control conditions or a nearby emergency cold storage facility from the available alternative transport resource pool, generates a relay transport or emergency storage scheduling plan, and pushes the plan to the relevant operator terminals.

8. The IoT-based online temperature monitoring system for the entire cold chain according to claim 1, characterized in that, The environmental sensor includes at least one of a humidity sensor, a door magnetic sensor, and a vibration sensor; the humidity sensor is used to monitor humidity data in the cold chain environment, the door magnetic sensor is used to monitor the opening and closing status and duration of the door in the space, and the vibration sensor is used to monitor impact vibration events exceeding a preset threshold during transportation; the data preprocessing module of the edge computing gateway uses a rotating door compression algorithm on the raw sensor data, and by setting a compression deviation threshold, approximates the continuously collected time-series data points as a line segment sequence, only uploading the endpoint data of the line segments, achieving a compression rate of at least 80% for the raw data, thereby reducing network bandwidth load.

9. The IoT-based online temperature monitoring system for the entire cold chain according to claim 1, characterized in that, The LoRa wireless communication module uses spread spectrum modulation technology to establish a wireless connection with the edge computing gateway in a metal enclosed car or cold storage environment, with a data packet loss rate of less than 0.3%. The BLE communication module is used to perform short-range data synchronization with a handheld mobile terminal during inspection or loading and unloading, enabling offline data retransmission and on-site parameter configuration.

10. The IoT-based online temperature monitoring system for the entire cold chain according to claim 1, characterized in that, When dynamically adjusting the temperature alarm threshold, the adaptive threshold algorithm specifically performs the following process: The anomaly detection module determines whether the current stage is warehousing, transportation, or distribution based on the position sensor and preset cold chain link logic. Read the standard temperature safety lower and upper limits of the transported goods from the cargo information synchronized from the cloud platform layer; The system receives the external ambient temperature collected by the environmental sensor in real time and calculates the slope of the change in the current external temperature within the sliding time window. When the absolute value of the slope exceeds the preset slope threshold, the upper and lower limits of the safety range are narrowed according to the magnitude of the slope by a proportional coefficient. The proportional coefficient is determined by a pre-calibrated cargo temperature rise sensitivity curve, so that the alarm threshold is narrowed in advance when the external heat load increases sharply, thus achieving early warning.

Images